BMC Pharmacology

BioMed CentralBMC Pharmacology

ss

Open AcceResearch articleBioinformatic characterizations and prediction of K+ and Na+ ion channels effector toxinsRima Solidagger1 Belhassen Kaabidagger13 Mourad Barhoumi1 Mohamed El-Ayeb2 and Najet Srairi-Abid2

Address 1Laboratory of Epidemiology and Ecology of Parasites Institut Pasteur de Tunis Tunis Tunisia 2Laboratory of Venom and Toxins Institut Pasteur de Tunis Tunis Tunisia and 3Research and Teaching Building Institut Pasteur de Tunis 13 Place Pasteur BP 74 1002 Belvedere-Tunis Tunisia

Email Rima Soli - rimafstyahoofr Belhassen Kaabi - belhassenkaabirnstn Mourad Barhoumi - m1barhoumiyahoocom Mohamed El-Ayeb - mohamedelayebpasteurrnstn Najet Srairi-Abid - najetabidpasteurrnstn

Corresponding author daggerEqual contributors

AbstractBackground K+ and Na+ channel toxins constitute a large set of polypeptides which interact withtheir ion channel targets These polypeptides are classified in two different structural groupsRecently a new structural group called birtoxin-like appeared to contain both types of toxins hasbeen described We hypothesized that peptides of this group may contain two conserved structuralmotifs in K+ andor Na+ channels scorpion toxins allowing these birtoxin-like peptides to be activeon K+ andor Na+ channels

Results Four multilevel motifs overrepresented and specific to each group of K+ andor Na+ ionchannel toxins have been identified using GIBBS and MEME and based on a training dataset of 79sequences judged as representative of K+ and Na+ toxins

Unexpectedly birtoxin-like peptides appeared to present a new structural motif distinct from thosepresent in K+ and Na+ channels Toxins This result supported by previous experimental datasuggests that birtoxin-like peptides may exert their activity on different sites than those targeted byclassic K+ or Na+ toxins

Searching the nr database with these newly identified motifs using MAST retrieved severalsequences (116 with e-value lt 1) from various scorpion species (test dataset) The filtering processleft 30 new and highly likely ion channel effectors

Phylogenetic analysis was used to classify the newly found sequences Alternatively classificationtree analysis using CART algorithm adjusted with the training dataset using the motifs and their2D structure as explanatory variables provided a model for prediction of the activity of the newsequences

Conclusion The phylogenetic results were in perfect agreement with those obtained by theCART algorithm

Our results may be used as criteria for a new classification of scorpion toxins based on functional motifs

Published 10 March 2009

BMC Pharmacology 2009 94 doi1011861471-2210-9-4

Received 31 July 2008Accepted 10 March 2009

This article is available from httpwwwbiomedcentralcom1471-221094

copy 2009 Soli et al licensee BioMed Central Ltd This is an Open Access article distributed under the terms of the Creative Commons Attribution License (httpcreativecommonsorglicensesby20) which permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Page 1 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

BackgroundThe most-studied components of scorpion venom arepolypeptides that recognize ion channels and receptors inexcitable membranes which are harmful to a variety oforganisms including human Two families of toxins thatinteract specifically with K+ and Na+ ion channels respec-tively [1] are the subject of intensive work in drug designand development [2-4]

These toxins have been classified according to species-spe-cificity (mammals insects and crustaceans) receptor tar-gets (K+ and Na+) their lengths (short or long chain)disulfide bonds arrangements [56] mechanism of actionand binding site α or β-like toxins [78]

Toxins that affect (modulate) Na+ channels which accountfor 1 to 10 of raw venom are long polypeptides with60ndash76 amino acid residues [7] reticulated and stabilizedby four disulphide bonds (S-S) [9-11] Three S-S arelocated in the molecular core and are conserved across thefamily while the fourth one is exposed on the molecularsurface and varies in position Considering this character-istic this disulfide bridge has been named wrapperdisulfide bridge [12]

K+ channels toxins are short-chain peptides (22ndash41 aminoacid residues) which are reticulated and stabilized bythree or four S-S [13] represent a minor component of theraw venom with the order from 005 to 01 [14] Inaddition these toxins have particular affinities and specif-icities for various K+ channel subfamilies [15]

Despite the great variation in the primary structures ofmany short and long toxins they share a common struc-tural three-dimensional (3D) conformation [7141617]

The current available online databases contain up to 800records of native and mutant toxin sequences enrichedwith binding affinity toxicity information and about 6503D structures Scorpion2 [18] and Tox-Prot [19] are twoexamples of comprehensive database available on theWeb

Recently a new structural group of toxins with 53ndash59amino acids and only 3 S-S called birtoxin-like peptideshave been characterized [561220-24] This structuralgroup contains peptides with similar sequences that showdifferences in activity Some peptides are active on Na+

channels [2023] while others are active or putativelyactive on both (K+ and Na+) channels [5622]

Given the above characteristics of this new group of tox-ins it is clear that classification of K+ or Na+ ion channelseffector toxins based on their lengths and the number ofS-S is not fully adequate to identify correctly the activity

of a given toxin The objectives of this study are1) identi-fication of signatures (motifs) associated with a givenactivity on the K+ andor Na+ channels 2) verifying thepresence of these motifs in the birtoxin-like family

Within this framework we planned to perform the fol-lowing steps sampling of toxins active on K+ andor Na+

ion channels and determination of the structural signa-ture corresponding to each type of the channels (K+ andNa+) effectors A statistical model (classification model)that uses these motifs and their secondary (2D) structureto predict the function of a given toxin was built

MethodsSequences preparation and highly similar sequences eliminationKey words K+ Na+ channel scorpion toxin were used tosearch the NCBI database [25] which is linked to swissprotpdb and embl among other databases for existing K+ andNa+ channel effectors (toxin sequences) Whole lengthand fragment sequences were included to insure maxi-mum coverage of these toxins with sequence information

All sequences were gathered according to their activitiesSequences that belong to the same group were put in thesame file in FASTA format Three groups were obtainedthe first contains toxins active on the Na+ channels thesecond contains toxins active on the K+ channels and thethird contains birtoxin-like peptides A perl script was usedto conduct batch PSI-BLAST [26] against the nr (non-redundant protein) database to search for similarity andto characterize these toxins

Multiple sequence alignmentMultiple sequences alignment was performed using Clus-talW [27] version 2 [28] Multiple alignments ofsequences for each group of toxins and for all sequencespooled together were carried out in order to characterizepotential conserved and variable areas for each sequencepartition as well as consensus sequences (centroides) Todate ClustalW is still the most popular alignment toolHowever recent methods in some cases offer signifi-cantly better alignment quality Thus this same stage wascarried out by another multiple alignments softwareCHAOS amp DIALIGN [29] which is based on a combina-tion of local and global alignment methodologies Theobtained results were compared with the former ones

In order to keep random noise at its lowest level all qual-ified sequences contained three or four S-S pattern Basedon these results highly similar sequences were removedto reduce potential bias on the motifs search (motifs con-served in each group) A subset of the originally foundsequences judged representative of the K+ andor Na+ ionchannel effectors were used as training dataset This sub-

Page 2 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

set reflects a trade-off between sequences that cover mostof scorpion species and sequences that are centroide in themultiple alignments However for the birtoxin-like groupall sequences were sampled

Motifs searchUsing the training dataset composed of the selected toxinsrepresentative of the ion channels blockers for K+ and Na+GIBBS [3031] version 205 and MEME [3233] version354 which is part of the Meta-MEME [34] package wereused to dig out conserved motif information for all con-sidered sequence partitions

We used GIBBS to discover the conserved motif(s) (withpre-fixed length which can contain(s) gaps) specific toeach group of ion channel effectors This same stage wasperformed by MEME which determines un-gapped mul-tilevel motifs (with no pre-fixed length) The most signif-icant results were compared and retained

Database search for new K+ and Na+ channel effector toxinsThe output of these programs (GIBBS and MEME) wasthoroughly investigated and then used as input in thenext step of the analysis

A profile using these motifs was systematically formedand used to search for sequences with this profile For thatpurpose MAST which is also part of the Meta-MEMEpackage was used to search for new K+ and Na+ channeleffector sequences (test dataset) against nr database Thee-value for a qualified sequence was set to 1 A perl scriptwas used to eliminate replications (identical to thesequences of the training dataset)

InterproscanInterPro [35] is an integrated resource for protein familiesdomains and functional sites which also integrates anumber of protein signature databases as well as theappropriate search tools for them The motifs found werethen searched for in the InterPro database using InterPro-Scan

2D-structure determinationThe distribution of the 2D structure in the motif region foreach sequence was studied in order to propose a relationlinking 2D structure to the function for each group of tox-ins The 2D structure of all the sequences (training andtest datasets) was determined based on the program PHD[36-38] using neural network approach and the Soft-berrys software PSSfinder [39] which uses Markov chainsprobabilistic model

Phylogenetic analysisPhylogenetic analysis was carried out to1) study thepotential evolutionary relations between the new

sequences and already known toxins 2) infer potentialcommon functions and 3) classify the new ones (withunknown functions) with respect to the other toxins withknown activities

Since multiple-alignment is at the base of any phylogenyalignments obtained from ClustalW results were used asentry for program PHYLIP [40] version 36 used to buildphylogenetic trees Distances were generated using Jones-Taylor-Thornton model Phylogenetic trees were con-structed using the neighbor-joining algorithm [41]

A construction of the phylogenetic trees for each ofsequence group considered was also performed Consen-sus trees were determined using the bootstrap method[42]

Statistical analysisCorrespondence analysis was performed and biplot [43]was drawn to illustrate the association between themotifs found and the activities of the toxins from thetraining dataset To test the strength of this association aχ2 test of independence of factors was used where p-valuewas computed using Monte-Carlo simulation [44] Thisprocedure is implemented in the R software version 27[45] and used in case where large sample theory is notvalid (many cell having values less than 5) The same testχ2 was used to measure association between the distribu-tion of the 2D structure (determined by PHD and PSS-finder) of the motifs and the toxins activities In order toarrange the toxins according to their activities (dependentvariable) and the motifs determined and their 2D struc-tures (structural variables) classification and regressiontrees (CART) [46] were built using the procedures imple-mented in the software Splus version 62 [47] The builtclassification trees (for each 2D program results) werethen used to predict the classes of the new sequences (testdataset) in a purely statistical way independent of thephylogeny results

ResultsUsing the key words K+ Na+ channel inhibitor toxinabout 700 sequences were found from the NCBI serverAmong them 495 are confirmed experimentally No ini-tial filtering was performed to avoid losing any potentialsignals All the 495 sequences were used to conduct batchPSI-BLAST one search iteration against the nr databaseThe E-value threshold required for sequence inclusion wasset to 10-5 After elimination of identical sequences andthose that do not have the S-S pattern all information oneach individual sequence of these potential channels tox-ins were retrieved Non-fragment sequences were dug outfrom the nr database and placed in an Excel file To pre-pare for the conserved region search by ClustalW allsequence partitions were placed in FASTA format in sepa-rate files Based on preliminary ClustalW results highly

Page 3 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

similar sequences were removed to reduce bias on con-served and functionally overrepresented motifs search

After careful examination of the resulting data 79sequences judged as representatives of the class of toxinsactive on K+ andor Na+ channels and confirmed experi-mentally were sampled These 79 toxins (training data-set) covered 18 scorpion species and distributed as follow27 toxins active on K+ channels 38 active on the Na+ chan-nels and 14 toxins belonging to the birtoxin-like groupClustalW re-applied on this training dataset revealed thefollowing

1) Eight cysteine residues implicated in the formation ofS-S in long toxins were conserved in all Na+ channel effec-tors except for CsI CsvI and CsvII (probably because oftheir loop between the second and third β strands whichis longer than the one for other Na+ channel scorpion tox-ins [48]) 2) Three cysteine residues were conserved in allK+ channel effectors 3) Five cysteine residues were con-served in the birtoxin-like group toxins 4) One motifADVPGNYPL was conserved in this group 5) For allsequences pooled together only 3 cysteine residues wereconserved and there is no conserved region (substring)

However ChaosDialign did not reveal any conservedregion for all sequence partitions

Motifs search resultsThe training dataset in its different partitions was used asinput to GIBBS and MEME to conduct motif search WhileGIBBS provides gapped or un-gapped motifs MEME iden-tifies only un-gapped motifs

GIBBS resultsBecause of the variable and the relatively small length ofthe considered sequences we opted for a motif length of10 and 20 Analysis of GIBBS output revealed that therewas one motif conserved in all Na+ channel effectorsexcept for CsV which is considered as structurally inter-mediate homologous to α and β toxins [49] A motif con-served for all K+ channel effectors except for two sequences(TsKapa and TsTxK-α) For the new group of birtoxin-likepeptides one motif was conserved For all sequencespooled together GIBBS did not identify any particularmotif of interest (present in all type of sequences) Similarresults were obtained with motif of length 20 Howeverthe motifs obtained using length 10 and 20 respectivelywere not nested and they did not overlap (Table 1)

MEME resultsUsing MEME and considering the relative diversity amongK+ and Na+ channel effectors and the length range of thesetoxins the maximum motif number was set to 6 (variedfrom 1 to 6) and maximum motif length was set to 20

Four motif runs were reasonable and provided the bestand the most significant distribution of motifs among thegroups of toxins Therefore we used this run as input tothe MAST and Meta-MEME programs These multilevelmotifs are in PSSM format and the consensus (most prob-able) strings are shown in the Table 2 and Figure 1

Motif-1 was conserved in 32 sequences all of them are α-type Na+ channel blockers Motif-2 was conserved in 5sequences all of them are β-type Na+ channel blockersMotif-3 was observed in 15 sequences among them 14toxins belong to the birtoxin-like group and only one isNa+ channel effector CsvII Motif-4 was observed in 16sequences where 14 of them are active on the K+ channelsand 2 are birtoxin-like peptides (birtoxin and ikitoxin) Theremaining 14 sequences (1 Na+ channel effector and 13K+ channel effectors) did not report statistically significantmatches to any of the consensus motifs described in table2 However these results do not exclude that thesesequences may contain other probabilistic variants of themotifs found Indeed 11 among the 13 K+ channel effec-tors were reported by GIBBS to have a conserved motif oflength 20 that overlap with the one identified by MEME(motif4) The comparison of the results obtained byMEME with those obtained by GIBBS showed that themotifs of length 20 determined by GIBBS were eitherincluded in the motifs determined by MEME or signifi-cantly overlapped with them However the resultsobtained by GIBBS are less significant in probabilisticterms and in the ability of differentiating among thegroups of toxins Therefore only motifs obtained byMEME will be adopted for the rest of this work

Identification of new K+ and Na+ channel effectorsUsing motifs obtained by MEME we conducted a MASTsearch against the nr database and we analyzed the resultsTo avoid missing any signal we have set the maximummotif number to 6 which may lead to an increase in thenumber of false positives due to model over fitting There-

Table 1 GIBBS Results

Motif Length Group Most Probable Motif

10 K+ AKCMNGKC-CY

10 Na+ ACYC--LPE-V-IW

10 birtoxin-like ADVPGNYPLD

20 K+ VPCT-SPQCI-PCK-A-M--GKCMNR

20 Na+ Q-LGRWGNACYC--LPD-VPIR--G-C

20 birtoxin-like VPGNYPLDKDGNTY-LELGEN

Page 4 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

fore as a remedy the e-value for a qualified sequence wasset to 1 which may minimize type-I error (false positives)One hundred and seventeen sequences of various scor-pion species were retrieved Not only all the identicalsequences but also the ones used as input (training data-set) from the search results were removed Thus 72sequences remained (test dataset) Then known K+ andNa+ channel effector sequences in the returned test datasetwere removed (42 sequences removed) In total 30 strongcandidates as potential new K+ orand Na+ channel effec-tor toxins were found among five scorpion species Detailsare listed in Table 3

Pattern and domain analysisPattern and domain analysis were further used to filter theresults All these 30 new K+ and Na+ channel effectorsequences have at least one of the 4 conserved motifs pre-viously found in this study We aligned these 30sequences and other known K+ and Na+ channel effectorsThese motifs were overrepresented among all thesesequences (newly identified and known toxins) thusindicating that the identified motifs in the study are usefulfor searching potential new K+ and Na+ channel effectorsfrom scorpions or plants

InterPro resultsSearching InterPro for the identified motifs showed thatmotif-1 corresponded to 2 hits PD000908 characteristic

of long chain scorpion toxins and PF00537 correspond-ing to Toxin_3 domain characteristic of Na+ channelinhibitors from scorpion or plants Motif-2 and motif-3did not correspond to any match Motif-4 correspondedto 2 hits PD003586 characteristic of short chain scorpiontoxins and PF00451 corresponding to Toxin_2 domainThe fact that 2 motifs among the 4 identified overlap withother already reported in InterPro is a corroboration ofour approach

2D resultsUsing PHD and PSSfinder all 2D structures of thesequences were determined PHD results showed that the2D structures displayed by motif-1 motif-2 motif-3 andmotif-4 were mainly β-sheet α-helix β-sheet and α-helixrespectively Using PSSfinder the 2D structure motif-1motif-3 and motif-4 were mainly β-sheet β-sheet α-helix or β-sheet respectively The 2D structure displayedby motif-2 was unidentified

Phylogenetic analysis resultsAll 152 sequences (training andor test datasets) includ-ing known and newly identified K+ andor Na+ channeltoxins found in our analysis were used to build phyloge-netic consensus trees Using the training dataset only (79sequences) the consensus phylogenetic tree obtainedrevealed 3 major groups or clusters (figure 2a)

Group G1 contains all sequences of α-type Na+ channeltoxins Group G2 contains toxins having some activitieson K+ channel This group can be further subdivided intotwo subgroups S1 containing toxins active on K+ channelonly and S2 involving birtoxin-like peptides Group G3contains β-type Na+ channel toxins The group S2 of bir-toxin-like peptides was situated between clusters S1 andG3 where S1 contains sequences active on K+ channel andG3 contains β-type Na+ channel toxins Therefore S2 canbe considered as a transition group from toxins active onK+ channel to those β-type active on Na+ channel (figure2b) This consensus tree presents the classical divergencebetween α and β Na+ channel toxins and it is consistentwith previous phylogenetic constructions made by[1449] even though they did not use the same set oftoxin sequences

Table 2 MEME Results

Motif Number Group Length Consensus (Most Probable) Motif

1 Na+ 20 GNACWCIELPDNVPIRIPGK2 Na+ 11 THLYEQAVVWP3 birtoxin-like 20 NYPLDSSDDTYLCAPLGENP4 K+ 20 KDAGMRFGKCMNRKCHCTPK

MEME motifs in logo formatFigure 1MEME motifs in logo format

Motif-1 Motif-2

Motif-3 Motif-4

Page 5 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

Table 3 Result of the activity prediction of the newly identified sequence using phylogenetic analysis and probabilities of classification by CART tree models adjusted with the motifs and their 2D structure

Access Number Inferred Activity by Phylogeny Classification Probability by CART PSSfinder (PHD)

Na+ K+ birtoxin-like

gb|AAF312971 Na+ 1 0 0

gb|AAA695571 Na+ 1 0 0

gb|AAT367461 Na+ 1 0 0

prf||0804800B Na+ 1 0 0

gb|AAD473761 K+ 0 1 0

gb|AAG396411 Na+ 1 0 0

prf||0804800A Na+ 1 0 0

pdb|1LQI| Na+ 1 0 0

gb|AAF348721 Na+ 1 0 0

gb|AAK068981 Na+ 1 0 0

gb|AAF314771 Na+ 1 0 0

gb|AAA695581 Na+ 1 0 0

gb|AAT367451 Na+ 1 0 0

gb|AAG005801 Na+ 1 0 0

gb|AAP336201 Na+ 1 0 0

gb|AAP343321 Na+ 1 0 0

gb|AAG096571 Na+ 1 0 0

gb|AAG396431 Na+ 1 0 0

gb|AAV642541 Na+ 1 0 0

gb|AAT522031 Na+ 1 0 0

emb|CAD605401 Na+ 1 0 0

gb|AAF294651 Na+ 1 0 0

gb|AAR080451 Na+ 1 0 0

pdb|1SEG|A Na+ 1 0 0

gb|AAR080441 Na+ 1 0 0

gb|AAB315281 Na+ 1 0 0

Page 6 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

To infer the function of the test dataset toxins (72sequences) all sequences from training and test datasetswere used to build a consensus phylogenetic tree see fig-ure 2 We examined the whereabouts of the sequencesfrom the test dataset with respect to the clusters (groups)defined by the training dataset Thus to perform a finalcheck on the reliability of the results obtained we exam-ined the whereabouts of the 42 sequences of the test data-set with known function in the output of PHYLIP withrespect to the clusters defined by the training dataset (79sequences) The results show that these sequences (42)with known activities fall within compatible functionallydefined clusters (clusters defined by toxins of the trainingdataset) thus supporting our results

The activities of the remaining 31 sequences were inferredin the same manner see table 3 and figure 1 We can seethat most newly identified peptides were grouped withsequences with known function toxins

Statistical analysis resultsThe association between generated motifs and the toxinsactivities was very significant simulated p-value lt 00001See biplot figure 3

Similarly and while the majority of the motifs displayeda β-sheet structure the association between motifs 2Dstructure distribution and the protein classes was evalu-ated with the same χ2 test and simulated p-value lt 0001Therefore 2D structures of the motifs were incorporatedas explanatory variables in the CART model

Classification trees were adjusted using the training data-set of 79 sequences Misclassification error rates for PHDand PSSfinder 2D based trees were less of 005 indicatinggood classifications The same models were used to pre-dict the activities of the 72 sequences identified by MASTThe 42 sequences with known activities were correctlyclassified while the remaining 30 new sequences of thetest dataset were assigned activities using CART modelThe predicted activities were in perfect concordance withthe results of the phylogeny This fact confirms the classi-fication tree models built using the training dataset as wellas the results of the phylogeny

Discussion and conclusionDue to the relative stability of K+ and Na+ channel effec-tors provided by 3 or 4 S-S they are used as tools for bio-logical investigation of the ion channel structure [5051]and represent potential candidates for use in medical andpharmacological applications

Toxins from certain structural family generally target thesame receiver with varied intensity However in generaleach K+ and Na+ channel effector is slightly different fromeach other which makes it possible to find suitable toxinsfor a specific application [52] This underlines the regularneed to identify new K+ and Na+ channel effector toxins

The analysis of the protein structures based on the con-served motifs is largely used and it is proven useful in theprediction of the protein functions [53-55]

While no single motif was conserved in all K+ and Na+

channel effector sequences this approach has permittedthe identification of 4 motifs overrepresented and specificto each sequence functionclass Therefore these motifsmay be used as a criteria for the classification of thesetoxin types in addition to the usual classification basedon sequence length and number of S-S

Due to the complex nature of the voltage gated Na+ chan-nel there are many regions of the protein that can beattacked and therefore it is not surprising that we foundtwo overrepresented motifs in Na+ channel toxins Inter-estingly both motifs (motif-1 and motif-2 for α and β tox-ins respectively) for Na+ channel toxins found in thisstudy overlap with sites (Hydrophobic face C-terminalregion and β2ndashβ3 strands) described as essential for Na+

channels binding [56-60] Motif-4 contains amino-acidsthat are important in K+ toxin activity [61-63]

It was expected that the birtoxin-like group (new group)would include both or at least one of the motifs character-istic of K+ and Na+ channel effectors (motif-1 motif-2 ormotif-4) However this was not the case and toxins ofthis new group did exhibit another overrepresented motif(motif-3) This finding suggests that toxins of the birtoxin-like group do not interact with the same functional sites as

emb|CAD605411 Na+ 1 0 0

gb|AAB214611 Na+ 1 0 0

gb|AAG396401 Na+ 1 0 0

gb|AAB214621 Na+ 1 0 0

PSSfinder based tree and (PHD based tree when different)

Table 3 Result of the activity prediction of the newly identified sequence using phylogenetic analysis and probabilities of classification by CART tree models adjusted with the motifs and their 2D structure (Continued)

Page 7 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

Page 8 of 11(page number not for citation purposes)

Consensus phylogenetic tree built using all sequences (training dataset and test dataset)Figure 2Consensus phylogenetic tree built using all sequences (training dataset and test dataset) (a) A Simplified phyloge-netic tree that displays three major clades labeled G1 G2 and G3 G2 is further subdivided into two groups S1 and S2 (b) Indi-vidual clades G1 S1 S2 and G3 Symbols ndash Training dataset sequences black circle active on Na+ channel black square active on K+ channel black triangle birtoxin-like ndash Test dataset sequences with known activity (_Na+) active on Na+ channel (_K+) active on K+ channel birtoxin-like _birtoxin-like

(a)

G3

G1

S1

S2

G2

(b)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

other long or short chains toxins This hypothesis is sup-ported by competitive binding experiments showing thatalthough KAaH1 (a member of the birtoxin-like group) isactive on Kv1 channels it did not displace iodinated α-DTx sKTX and CTx (Kv1 channels blockers) from ratbrain synaptosomes (Abid unpublished data) SimilarlyAaBTX-L1 which is active on Na+ channel (and also amember of the birtoxin-like group) did not compete with125I-CssIV (Na+ channel blocker) Moreover no competi-tion was observed either with 125I-sKTX (K+ blocker) orwith 125I-Apamine (SKCa channels blocker) on their recep-tor sites on rat brain synaptosomes [23] Moreover bir-toxin ikitoxin dortoxin and alitoxin do not enhance thebinding of [H]BTX to rat brain synaptosomes which isnot consistent with their action on voltage-gated Na+ cur-rent [20]

birtoxin and ikitoxin were found to contain 2 motifs motif-3 specific to birtoxin-like peptides group and motif-4which is overrepresented in K+ channel toxins This sug-gests that birtoxin and ikitoxin could interact with K+ chan-nel through this motif However this hypothesis needs tobe verified by testing these toxins on K+ channels

The result obtained by the CART algorithm based on theidentified motifs and their 2D structures provided anidentical classification to that obtained by the phylogenyTherefore the knowledge about the motifs and their 2Dare sufficient to infer the activity of a given toxin This con-

clusion is supported by the fact that binding sites of toxinsto their channels target are generally situated on the α-helix or β-sheet 2D structures [226061]

The search for conserved motifs and the phylogeneticanalyses enabled us to find common characteristics toeach protein family and thus to predict the structure andthe function of the new protein sequences These motifsallowed us to find sequences that we were not able to findwith the classical criteria of toxin length and S-S number

All 30 identified potential K+ and Na+ channel effector tox-ins possess the overrepresented motifs specific to eachgroup of K+ andor Na+ channel effector and the S-Sdomain While the e-value set for MAST was equal to onewhich is a quite stringent criterion it led to the discoveryof very significant motifs with potential biological func-tion (modulating specific ion channels) and providedhigh sensitivity and minimized the false positives withrespect to MAST use

Moreover the phylogeny has shown that all the newlyidentified potential K+ and Na+ channel effectors wereclosely grouped to other known toxins The newsequences were situated inside the groups limited by tox-ins with known functions

In addition the birtoxin-like peptides (S2) were locatedbetween cluster containing toxins active on K+ channels(S1) and cluster containing Na+ channel β-type toxinsTherefore S2 sequences can be considered as a transitiongroup putatively active on both channels (K+ and Na+)Indeed this fact explains the presence of both types (K+and Na+) of toxins in this group The challenge will be tofind in the birtoxin-like (growing in size) group which arethe signatures responsible for K+ or Na+ channel modula-tion Unfortunately this group does not contain till nowsufficient sequences and was not fully characterized withrespect to K+ and Na+ channels activities to make a consist-ent conclusion

Future work will deal with a finalization of the identifiedmotifs as to discern the exact number of residues loca-tion and implication for the toxin activities DockingStudy and building biophysical models that incorporatethese motifs and model the interaction with their targetswill be of great use

Competing interestsThe authors declare that they have no competing interests

Authors contributionsRS performed the sequences extraction produced thetraining data and performed the analysis BK performedthe statistical analysis and part of the sequence analysis

Biplot of the correspondence analysis distribution of the motifs and the toxin functionsFigure 3Biplot of the correspondence analysis distribution of the motifs and the toxin functions

Comp1

Com

p2

-05 00 05

-10

-05

00

05

10

-40 -20 0 20 40

-40

-20

020

40

Motif1

Motif2Na+Motif3

birtoxin-like

Motif4

K+

Page 9 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

MB studied the biological significance of the results andperformed a critical review of the manuscript ME co-directed this work and performed a critical review of themanuscript NSA supervised and co-directed the studyand performed a critical review of the manuscript Allauthors contributed significantly in the drafting of themanuscript All authors have read and approved the man-uscript

AcknowledgementsThis study received financial support from the Secretariat of the State for Scientific Research Technology and Competencies Development in Tuni-sia through funding of Research Program Contract (2004ndash2008) for Institut Pasteur de Tunis We address warm thanks to Pr Alan L Harvey Dr Louize Young (Department of Physiology and Pharmacology and Strath-clyde Institute for Drug Research University of Strathclyde Glasgow UK) Dr Marie-France Martin Eauclaire (Laboratoire dIngeacutenierie des Proteines Faculteacute de Meacutedecine de Marseille France) for displacement tests of KAah1 and KAah2 on synaptosome rat brain

We are grateful to anonymous reviewers for their comments and to Drs Dhafer Laouini and Elyes Zhioua for having reviewed the manuscript

References1 Catterall WA Cestegravele S Yarov-Yarovoy V Yu FH Konoki K

Scheuer T Voltage-gated ion channels and modifier toxinsVoltage-gated ion channels and effector toxins Toxicon 200749(2)124-141

2 Lewis RJ Ion channel toxins and therapeutics from cone snailvenoms to ciguatera Ther Drug Monit 2000 2261-64

3 Gazarian KG Gazarian T Hernandez R Possani LD Immunologyof scorpion toxins and perspectives for generation of anti-venom vaccines Vaccine 2005 233357-3368

4 Priest BT Kaczorowski GJ Blocking sodium channels to treatneuropathic pain Expert Opin Ther Targets 2007 11(3)291-306

5 Zhu S Gao B Molecular characterization of a possible progen-itor sodium channel toxin from the Old World scorpion Mes-obuthus martensii FEBS Lett 2006 580(25)5979-5987

6 Zhu S Gao B Molecular characterization of a new scorpionvenom lipolysis activating peptide Evidence for disulfidebridge-mediated functional switch of peptides FEBS Lett 2006580(30)6825-36

7 Possani LD Becerril B Delepierre M Tytgat J Scorpion toxins spe-cific for Na+ channels Eur J Biochem 1999 264287-300

8 Priest BT Blumenthal KM Smith JJ Warren VA Smith MM ProTx-I and ProTx-II Modifiers of voltage-gated sodium channelsToxicon 2007 49(2)194-201

9 Kopeyan C Martinez G Lissitzky S Miranda F Rochat H Disulfidebonds of toxin II of the scorpion Androctonus australis Hec-tor Eur J Biochem 1974 47483-489

10 Darbon H Zlotkin E Kopeyan C van Rietschoten J Rochat H Cov-alent structure of the insect toxin of the North African scor-pion Androctonus australis Hector Int J Pept Protein Res 198220(4)320-330

11 Martin-Eauclaire MF Ceacuteard B Ribeiro AM Diniz CR Rochat HBougis PE Biochemical pharmacological and genomic char-acterisation of Ts IV an alpha-toxin from the venom of theSouth American scorpion Tityus serrulatus FEBS Lett 1994342(2)181-184

12 Inceoglu B Lango J Wu J Hawkins P Southern J Hammock BD Iso-lation and characterization of a novel type of neurotoxicpeptide from the venom of the South African scorpionParabuthus transvaalicus (Buthidae) Eur J Biochem 20012685407-5413

13 Rodriacuteguez de la Vega RC Possani LD Current views on scorpiontoxins specific for K+ channels Toxicon 2004 43(8)865-75

14 Legros C Martin-Eauclaire MF [Scorpion toxins] C R SeancesSoc Biol Fil 1997 191(3)345-80 French

15 Legros C Pollmann V Knaus HG Farrell AM Darbon H Bougis PEMartin-Eauclaire MF Pongs O Generating a high affinity scor-pion toxin receptor in KcsA-Kv13 chimeric potassium chan-nels J Biol Chem 2000 27516918-16924

16 Bontems F Roumestand C Gilquin B Meacutenez A Toma F Refinedstructure of charybdotoxin common motifs in scorpion tox-ins and insect defensins Science 1991 2541521-1523

17 Bontems F Roumestand C Boyot P Gilquin B Doljansky Y MenezA Toma F Three-dimensional structure of natural charyb-dotoxin in aqueous solution by 1H-NMR Charybdotoxinpossesses a structural motif found in other scorpion toxinsEur J Biochem 1991 19619-28

18 Tan PT Veeramani A Srinivasan KN Ranganathan S Brusic VSCORPION2 a database for structure-function analysis ofscorpion toxins Toxicon 2006 47(3)356-63

19 Jungo F Bairoch A Tox-Prot the toxin protein annotation pro-gram of the Swiss-Prot protein knowledgebase Toxicon 200545(3)293-301

20 Inceoglu AB Hayashida Y Lango J Ishida AT Hammock BD A singlecharged surface residue modifies the activity of ikitoxin a -type Na+ channel toxin from Parabuthus transvaalicus Eur JBiochem 2002 2695369-5376

21 Inceoglu B Lango J Pessah IN Hammock BD Three structurallyrelated highly potent peptides from the venom ofParabuthus transvaalicus possess divergent biological activ-ity Toxicon 2005 45(6)727-733

22 Srairi-Abid N Guijarro JI Benkhalifa R Mantegazza M Cheikh A BenAissa M Haumont PY Delepierre M El Ayeb M A new type ofscorpion Na+ channel-toxin-like polypeptide active on K+

channels Biochem J 2005 388(2)455-6423 Martin-Eauclaire MF Ceard B Bosmans F Rosso JP Tytgat J Bougis

PE New Birtoxin analogs from Androctonus australisvenom Biochem Biophys Res Commun 2005 333(2)524-530

24 Caliskan F Garciacutea BI Coronas FI Batista CV Zamudio FZ PossaniLD Characterization of venom components from the scor-pion Androctonus crassicauda of Turkey peptides andgenes Toxicon 2006 48(1)12-22

25 Wheeler DL Barrett T Benson DA Bryant SH Canese KChetvernin V Church DM Dicuccio M Edgar R Federhen S FeoloM Geer LY Helmberg W Kapustin Y Khovayko O Landsman DLipman DJ Madden TL Maglott DR Miller V Ostell J Pruitt KDSchuler GD Shumway M Sequeira E Sherry ST Sirotkin K SouvorovA Starchenko G Tatusov RL Tatusova TA Wagner L Yaschenko EDatabase resources of the National Center for Biotechnol-ogy Information Nucleic Acids Res 2008D13-21

26 Altschul SF Madden TL Schaumlffer AA Zhang J Zhang Z Miller W Lip-man DJ Gapped BLAST and PSI-BLAST a new generation ofprotein database search programs Nucleic Acids Res 199725(17)3389-402

27 Thompson JD Higgins DG Gibson TJ CLUSTALW Improvingthe sensitivity of progressive multiple sequence alignmentthrough sequence weighting position-specific gap penaltiesand weight matrix choice Nucleic Acids Res 1994 224673-4680

28 Larkin MA Blackshields G Brown NP Chenna R McGettigan PAMcWilliam H Valentin F Wallace IM Wilm A Lopez R ThompsonJD Gibson TJ Higgins DG ClustalW2 and ClustalX version 2Bioinformatics 2007 23(21)2947-2948

29 Brudno M Steinkamp R Morgenstern B The CHAOSDIALIGNWWW server for multiple alignment of genomic sequencesNucleic Acids Res 2004W41-4

30 Thompson W Rouchka EC Lawrence CE Gibbs Recursive Sam-pler finding transcription factor binding sites Nucleic Acids Res2003 313580-3585

31 Thompson WA Newberg LA Conlan S McCue LA Lawrence CEThe Gibbs Centroid Sampler Nucleic Acids Res 2007W232-7

32 Bailey TL Elkan C Unsupervised learning of multiple motifs inbiopolymers using expectation maximization Machine Learn-ing 1995 2151-80

33 Bailey TL Elkan C The value of prior knowledge in discoveringmotifs with MEME In Proceedings of the Third International Confer-ence on Intelligent Systems for Molecular Biology AAAI Press 199521-29

34 Grundy WN Bailey TL Elkan CP Baker ME Meta-MEME Motif-based Hidden Markov Models of Biological Sequences Com-put Appl Biosci 1997 13(4)397-406

35 Mulder NJ Apweiler R Attwood TK Bairoch A Bateman A Binns DBork P Buillard V Cerutti L Copley R Courcelle E Das U Daugh-

Page 10 of 11(page number not for citation purposes)

BMC Pharmacology 2009 94 httpwwwbiomedcentralcom1471-221094

Publish with BioMed Central and every scientist can read your work free of charge

BioMed Central will be the most significant development for disseminating the results of biomedical research in our lifetime

Sir Paul Nurse Cancer Research UK

Your research papers will be

available free of charge to the entire biomedical community

peer reviewed and published immediately upon acceptance

cited in PubMed and archived on PubMed Central

yours mdash you keep the copyright

Submit your manuscript herehttpwwwbiomedcentralcominfopublishing_advasp

BioMedcentral

erty L Dibley M Finn R Fleischmann W Gough J Haft D Hulo NHunter S Kahn D Kanapin A Kejariwal A Labarga A Langendijk-Genevaux PS Lonsdale D Lopez R Letunic I Madera M Maslen JMcAnulla C McDowall J Mistry J Mitchell A Nikolskaya ANOrchard S Orengo C Petryszak R Selengut JD Sigrist CJ ThomasPD Valentin F Wilson D Wu CH Yeats C New developments inthe InterPro database Nucleic Acids Res 2007D224-8

36 Rost B Sander C Combining evolutionary information andneural networks to predict protein secondary structure Pro-teins 1994 19(1)55-72

37 Rost B PHD predicting one-dimensional protein structureby profile-based neural networks Methods Enzymol 1996266525-39

38 Rost B Yachdav G Liu J The PredictProtein Server Nucleic AcidsResearch 2004W321-W326

39 PSSfinder SoftBerry website [httplinux1softberrycomberryphtml]

40 Felsenstein J Inferring Phylogenies Sinauer Associates Sunder-land Massachusetts 2004

41 Saitou N Nei M The neighbor-joining method a new methodfor reconstructing phylogenetic trees Mol Biol Evol 19874406-425

42 Zharkikh A Li W-H Estimation of confidence in phylogeny thecomplete-and-partial bootstrap technique Mol Phylogenet Evol1995 4(1)44-63

43 Gabriel KR Odoroff CL Biplots in biomedical research Statisticsin Medicine 1990 9469-485

44 Hope ACA A simplified Monte Carlo significance test proce-dure J Roy Statist Soc 1968 30582-598

45 R version 27 the R Project for Statistical Computing homepage [httpwwwr-projectorg]

46 Breiman L Fridman JH Olshen RA Stone CJ Classification andRegression Trees In Statistical Models in S Edited by Chambers JMHastie TJ Belmont CA Wadsworth International Group 1984414

47 Insightful Corp S-PLUS software S-PLUSreg 62 for WindowsPROFESSIONAL EDITION Seattle USA 2003

48 Cook WJ Zell A Watt DD Ealick SE Structure of variant 2 scor-pion toxin from Centruroides sculpturatus Ewing Protein Sci-ence 2002 11479-486

49 Rodriacuteguez de la Vega RC Possani LD Overview of scorpion tox-ins specific for Na+ channels and related peptides biodiver-sity structure-function relationships and evolution Toxicon2005 46831-844

50 Desai SA Targeting ion channels of Plasmodium falciparum-infected human erythrocytes for antimalarial developmentCurr Drug Targets Infect Disord 2004 4(1)79-86

51 Rogers M Tang L Madge DJ Stevens EB The role of sodium chan-nels in neuropathic pain Semin Cell Dev Biol 2006 17(5)571-581

52 Gordon D Gurevitz M The selectivity of scorpion a-toxins forsodium channel subtypes is determined by subtle variationsat the interacting surface Toxicon 2003 41125-128

53 Cardoso FC Pinho JM Azevedo V Oliveira SC Identification of anew Schistosoma mansoni membrane-bound proteinthrough bioinformatic analysis Genet Mol Res 20065(4)609-618

54 Hedlund J Cantoni R Baltscheffsky M Baltscheffsky H Analysis ofancient sequence motifs in the H-PPase family FEBS J B 2006273(22)5183-5193

55 Liu F Baggerman G Schoofs L Wets G Uncovering conservedpatterns in bioactive peptides in Metazoa Peptides 200627(12)3137-3153

56 Sampieri F Habersetzer-Rochat C Structure-function relation-ships in scorpion neurotoxins Identification of the supperre-active lysine residue in toxin I of Androctonus australisHector Biochim Biophys Acta 1978 535(1)100-109

57 Srairi-Abid N Mansuelle P Mejri T Karoui H Rochat H Sampieri FEl Ayeb M Purification characterization and molecular mod-elling of two toxin-like proteins from the Androctonus aus-tralis Hector venom Eur J Biochem 2000 267(17)5614-5620

58 Sun YM Bosmans F Zhu RH Goudet C Xiong YM Tytgat J WangDC Importance of the conserved aromatic residues in thescorpion alpha -like toxin BmK M1 the hydrophobic surfaceregion revisited J Biol Chem 2003 278(26)24125-24131

59 Cohen L Karbat I Gilles N Froy O Corzo G Angelovici R GordonD Gurevitz M Dissection of the functional surface of an anti-insect excitatory toxin illuminates a putative hot spot com-

mon to all scorpion beta-toxins affecting Na+ channels J BiolChem 2004 2798206-8211

60 Cohen L Karbat I Gilles N Ilan N Benveniste M Gordon D Gure-vitz M Common features in the functional surface of scorpionbeta-toxins and elements that confer specificity for insectand mammalian voltage-gated sodium channels J Biol Chem2005 280(6)5045-53

61 Dauplais M Lecoq A Song J Cotton J Jamin N Gilquin B Roume-stand C Vita C de Medeiros CL Rowan EG Harvey AL Meacutenez AOn the convergent evolution of animal toxins Conservationof a diad of functional residues in potassium channel-block-ing toxins with unrelated structures J Biol Chem 1997272(7)4302-4309

62 Regaya I Beeton C Ferrat G Andreotti N Darbon H De Waard MSabatier JM Evidence for domain-specific recognition of SKand Kv channels by MTX and HsTx1 scorpion toxins J BiolChem 2004 279(53)55690-55696

63 Mouhat S Visan V Ananthakrishnan S Wulff H Andreotti N Griss-mer S Darbon H De Waard M Sabatier JM K+ channel types tar-geted by synthetic OSK1 a toxin from Orthochirusscrobiculosus scorpion venom Biochem J 2005 38595-104

Page 11 of 11(page number not for citation purposes)